I am applying supervised classification on landsat 5 images. But there is certian problem in obtaining the results. As you can see in the picture that there are settlements along the stream but if I see on google map it doesn't show the settlements along the river.
I suggest:
Enter the map to the ArcGIS software and classified the maps, then, re-integrate the classes.
Validate to clean the imagery of noisy data with a spectral classification algorithm
You can use random sampling method in google earth . Use same temporal range google earth image for that.First you can mark LULC types ( water bodies, built up areas,vegetation areas...) as points or polygons using google earth. Then you can export the data as training samples for supervised classification.
After that you can do an accuracy assessment.
- Badges
- Science method